[Back]


Diploma and Master Theses (authored and supervised):

B. Flores:
"Pre-Distortion Algorithms Implemented in Fixed-Point Arithmetic";
Supervisor: R. Dallinger, M. Rupp; Institute of Telecommunications, 2015.



English abstract:
Nowadays, wireless communications systems are required to provide high data-rates
with high quality. In order to achieve this, spectrally e cient transmission techniques
are employed which rely on signals with large envelope
uctuations. Moreover, due to
power e ciency demands power ampli ers have to work close to their saturation region.
Unfortunately, their resulting nonlinear behaviour introduces nonlinear distortions. By
this, on the one hand the transmitted signal is degraded, on the other hand, it causes
spectral widening beyond the channel bandwidth, and consequently interference with
neighbouring transmission channels.
Digital pre-distortion is a technique used to compensate the distortions introduced by the
power ampli er, so that the overall system operates as a linear yet e cient amplifying
stage. This solution reduces the transmission unit size and allows for cutting energy
costs, especially if combined with other linearization techniques. As the pre-distorter
has to predict the nonlinearity introduced by the power ampli er, pre-distortion can be
considered a behavioural modeling problem.
In this thesis, we consider several pre-distortion schemes found in literature that are
based on behavioural modeling. Starting with the memoryless polynomial model, we
move on to the general but computationally expensive truncated Volterra series and,
nally end up with the decomposed piecewise Volterra series proposed by Zhu in [1] that
allow to reduce the computational complexity by selectively pruning of the truncated
Volterra series. The main goal of this work is to evaluate the xed-point implementation
of the algorithms. In order to do so the algorithms are implmented in MATLAB in
xed-point arithmetic, as well as in
oating-point arithmetic; where the latter is used
as reference for a comparison of performance. In addition, a detailed review of the
theory is presented in this work. The algorithms are evaluated with a nonlinear reference
model: a saleh model for the memoryless case and a hammerstein model for the memory
cases. Simulation results show that the decomposed piecewise Volterra model employing
the dynamic deviation reduction-based Volterra model as sub-model outperforms the
traditional models.

German abstract:
Nowadays, wireless communications systems are required to provide high data-rates
with high quality. In order to achieve this, spectrally e cient transmission techniques
are employed which rely on signals with large envelope
uctuations. Moreover, due to
power e ciency demands power ampli ers have to work close to their saturation region.
Unfortunately, their resulting nonlinear behaviour introduces nonlinear distortions. By
this, on the one hand the transmitted signal is degraded, on the other hand, it causes
spectral widening beyond the channel bandwidth, and consequently interference with
neighbouring transmission channels.
Digital pre-distortion is a technique used to compensate the distortions introduced by the
power ampli er, so that the overall system operates as a linear yet e cient amplifying
stage. This solution reduces the transmission unit size and allows for cutting energy
costs, especially if combined with other linearization techniques. As the pre-distorter
has to predict the nonlinearity introduced by the power ampli er, pre-distortion can be
considered a behavioural modeling problem.
In this thesis, we consider several pre-distortion schemes found in literature that are
based on behavioural modeling. Starting with the memoryless polynomial model, we
move on to the general but computationally expensive truncated Volterra series and,
nally end up with the decomposed piecewise Volterra series proposed by Zhu in [1] that
allow to reduce the computational complexity by selectively pruning of the truncated
Volterra series. The main goal of this work is to evaluate the xed-point implementation
of the algorithms. In order to do so the algorithms are implmented in MATLAB in
xed-point arithmetic, as well as in
oating-point arithmetic; where the latter is used
as reference for a comparison of performance. In addition, a detailed review of the
theory is presented in this work. The algorithms are evaluated with a nonlinear reference
model: a saleh model for the memoryless case and a hammerstein model for the memory
cases. Simulation results show that the decomposed piecewise Volterra model employing
the dynamic deviation reduction-based Volterra model as sub-model outperforms the
traditional models.

Keywords:
LMS Algorithm, Fixed-point


Electronic version of the publication:
http://publik.tuwien.ac.at/files/PubDat_242282.pdf



Related Projects:
Project Head Markus Rupp:
Signal and Information Processing in Science and Engineering II: Theory and Implementation of Distributed Algorithms


Created from the Publication Database of the Vienna University of Technology.